Merge pull request #2768 from langbot-app/fix/sqlite-startup-wal-checkpoint

fix(sqlite): reclaim startup WAL and ship patched SQLite

src/langbot/pkg/persistence/databases/sqlite.py

@@ -6,6 +6,19 @@ import sqlalchemy.ext.asyncio as sqlalchemy_asyncio
 from .. import database
 
 
+WAL_RETAIN_BYTES = 64 * 1024 * 1024
+
+
+def _configure_connection(dbapi_connection, _connection_record):
+    # This limit applies when a WAL resets, not while a transaction or reader
+    # still needs its frames. Keep SQLite's automatic checkpoint policy.
+    cursor = dbapi_connection.cursor()
+    try:
+        cursor.execute(f'PRAGMA journal_size_limit={WAL_RETAIN_BYTES}')
+    finally:
+        cursor.close()
+
+
 @database.manager_class('sqlite')
 class SQLiteDatabaseManager(database.BaseDatabaseManager):
     """SQLite database manager"""
@@ -14,6 +27,7 @@ class SQLiteDatabaseManager(database.BaseDatabaseManager):
         db_file_path = self.ap.instance_config.data.get('database', {}).get('sqlite', {}).get('path', 'data/langbot.db')
         engine_url = f'sqlite+aiosqlite:///{db_file_path}'
         self.engine = sqlalchemy_asyncio.create_async_engine(engine_url)
+        sqlalchemy.event.listen(self.engine.sync_engine, 'connect', _configure_connection)
         # Reader transactions must not block independent plugin callback commits.
         # Configure the persistent journal before serving requests, not per write.
         async with self.engine.connect() as connection:

src/langbot/pkg/persistence/mgr.py

@@ -198,6 +198,12 @@ class PersistenceManager:
             return
 
         await self._initialize_managed_schema()
+        from .sqlite_startup import checkpoint_after_migration
+
+        if engine.dialect.name == 'sqlite':
+            self.ap.logger.info('Completing SQLite startup WAL checkpoint before starting producers...')
+            await checkpoint_after_migration(engine)
+            self.ap.logger.info('SQLite startup WAL checkpoint completed.')
         await self._enable_sqlite_foreign_keys_after_migration()
         from .resource_urls import install
 

src/langbot/pkg/persistence/sqlite_startup.py

@@ -0,0 +1,27 @@
+"""Finish SQLite WAL maintenance before application producers start."""
+
+STARTUP_BUSY_TIMEOUT_MS = 60_000
+
+
+async def checkpoint_after_migration(engine):
+    """Require a complete WAL checkpoint, including on a retry at schema head."""
+    if engine.dialect.name != 'sqlite':
+        return
+    async with engine.connect() as connection:
+        connection = await connection.execution_options(isolation_level='AUTOCOMMIT')
+        if (await connection.exec_driver_sql('PRAGMA journal_mode')).scalar_one().lower() != 'wal':
+            return
+        timeout = (await connection.exec_driver_sql('PRAGMA busy_timeout')).scalar_one()
+        try:
+            await connection.exec_driver_sql(f'PRAGMA busy_timeout={STARTUP_BUSY_TIMEOUT_MS}')
+            busy, frames, checkpointed = (
+                await connection.exec_driver_sql('PRAGMA main.wal_checkpoint(TRUNCATE)')
+            ).one()
+            if (busy, frames, checkpointed) != (0, 0, 0):
+                raise RuntimeError(
+                    'SQLite startup WAL checkpoint did not complete: '
+                    f'busy={busy}, frames={frames}, checkpointed={checkpointed}. '
+                    'Stop other database users and restart LangBot to retry.'
+                )
+        finally:
+            await connection.exec_driver_sql(f'PRAGMA busy_timeout={timeout}')